/watch — Claude watches a video

You don't have a video input; this skill gives you one. A Python script downloads the video, extracts frames as JPEGs (one per detected shot via scene-change), gets a timestamped transcript (native captions first, then Whisper API as fallback), runs editorial pacing metrics, and microscopes the first 10 seconds at higher density. You then Read each frame path to see the images, combine them with the transcript to answer the user, fill the structured report.md, and offer to ingest the analysis into Taoufik's Second Brain.

What v2 does differently

Scene-change frame sampling — one frame per detected shot instead of uniform ticks. Cuts the frame budget on long videos while capturing every transition.
Editorial pacing metrics — cuts/min, mean shot length, motion (when available). Lets you reason about pacing the way an editor does.
Hook microscope — first 10s auto-runs at 2 fps + word-level Whisper. The single most leveraged 10 seconds of any video deserves dense treatment.
Structured report.md — every watch emits an ingest-shaped report at <workdir>/report.md with TL;DR, key moments, hook breakdown, editorial profile, quotable moments, entities, concepts, and transcript. Narrative sections are emitted as  markers — you fill them in before offering ingest.
Step 4.5 — Ingest gate — after answering the user, you ask once: "Want to ingest this into your Obsidian vault?" If yes, and a vault is detected, you read $VAULT_DIR/CLAUDE.md (if it exists) and run that vault's Ingest op against the report.

None of the above add new dependencies — pure ffmpeg + stdlib + the existing Whisper backend.

Configuration — finding the user's Obsidian vault

Steps 4.4 and 4.5 stage the report inside an Obsidian vault so the user can read it where they read everything else. Resolve the vault directory in this order — first hit wins, and the result is what $VAULT_DIR refers to everywhere below: